18 INNOVATION&TECHTODAY^ |^ 2020 YEAR-END ISSUE
DEPARTMENTS / Artificial Intelligence
A Brave New World Exploring the Ethics of AI
By Gretchen Philyaw
In the 1970s, Sneaky Pete was
the first chess-playing
computer, paving the way for
IBM’s Deep Blue — the first
machine using Artificial
Intelligence (AI) to become a
Chess Grandmaster. And of
course, ultimately, Alpha Zero
has transcended all. AI has
come a long way from those
rudimentary chess matches,
and the power of AI mandates a
set of ethics be established to
guide us into the future.
Machine learning is a
common term associated with
AI. In simple terms, software
developers write algorithms to
sort through data. The
algorithm has rules on how the
data is sorted or analyzed.
Based on the algorithm, the
machine can make decisions.
These decisions are rule-
based and they consider a lot of
data the human brain cannot
process. Over time, the
computer revises earlier
decisions based on new data.
The advantage of AI is that
with more data, decisions get
better.
The ethical dilemmas are
subtle regarding AI. Algorithms
are written to sort data and
make decisions. These
decisions can determine who
gets a bank loan, who gets
audited, or who receives a
college scholarship based on
thousands of applications.
There are many more examples
ofhowAIisused.Bias
The first ethical question is whether the algorithm
was biased against race, sex, or any other variable. If
so, the outcomes or decisions cannot be used. My
vision is that applications using AI will be
monitored just like the IRS monitors tax returns. A
formal reporting body will review the output of AI
logic to validate that there are no biases, and the
algorithm rules were followed correctly. This
validation will get reported in some official manner.Privacy
With data, there must be rules about privacy.
Consumers need to know when their information is
being used in any way. Consent must be given, and
any time AI is used to make a decision, consumers
or customers need to know how this decision was
derived. This privacy agreement cannot be hidden
in the fine print.Security
Security is a no brainer if data used in the AI
process is corrupted, hacked, or misused. To
ethically use AI, data integrity is paramount. If users
of AI cannot secure the data, how can they trust the
decision output?Algorithm Integrity
The final pillar of the ethics debate is the
legitimacy of the algorithm. Testing has to be
conducted using real-world data to determine if the
algorithm followed the rules set. Humans develop
the algorithm. Humans should also be able to verify
the algorithm was followed. Outputs cannot be
random.
Artificial Intelligence is a powerful tool. Data
analytics combined with immense amounts of data
pulled into an algorithm designed to make critical
decisions will be a big part of the future.
International governments will have to work
together to make the rules on how AI is harnessed
ethically. AI will create fantastic opportunities for
those willing to explore the future. ■About the Author: Gretchen
Philyaw is President of
Magenta Technologies LLC,
an Atlanta-based industrial
automation company and
founder of USA Loves
Manufacturing , an organization promoting and
advocating USA manufacturing and technology.
Photos: iStockphoto.com